Cooling

The building’s air conditioning doesn’t run on electricity the way a normal air conditioner does — it runs on heat. An absorption chiller sits in its own bay right next to the fuel cell, and it’s fired directly by the fuel cell’s exhaust, which leaves at 356–384°C. That’s the whole reason this bay is positioned where it is: everywhere else in the plant run, position is flexible, but cooling has to sit next to the fuel cell because hot exhaust gas is its actual fuel.

The unit itself is rated at 100 tons of cooling, though the real figure the plant can count on is closer to 75 tons once you account for how hot the fuel cell’s exhaust actually runs. It measures 2.66 × 1.86 × 2.20 m and weighs 5.4 tonnes.

Here’s the part that surprises most people: making cooling this way produces more heat than it removes. For every 1 kW of cooling the chiller delivers, it has to get rid of 1.7 kW of heat — a built-in cost of running on heat instead of electricity. Across the whole plant, that adds up to 727 kW that has to go somewhere, and it’s split three ways: 163 kW is absorbed by the pool, which just quietly warms up a few degrees over a hot afternoon; 72 kW escapes through the building’s outer walls; and the largest share, 491 kW, is pushed out through roof-mounted coils — 33 square metres of them, moving 74 cubic metres of air every second, using large, slow fans that only draw 17 kW total.

Those fans matter more than their size suggests. Because fan and pump power follows a cube law, slowing a fan to 40% speed only costs about 6.5% of its full power — so cooling can scale down cheaply on a mild day without wasting energy the way a fixed-speed system would.

Where the design runs into a real fit problem

Today’s chiller is 3,728 mm long. The bay it needs to fit into, once every other system is compressed to save space, is only 3,750 mm — leaving almost no room for the 2,400 mm of clearance a technician needs to service it. This is one of four machines in the whole plant run that simply doesn’t fit once the building’s five-year target footprint is applied. It’s not a hypothetical concern; it’s a real spec on a real bay that a real unit currently overflows.

Governed by Bureau 23.